آیا برنامه ریزی استفاده از زمین به اقتصاد منطقه ای ، شکل می دهد ؟ تجزیه و تحلیل همزمان تامین مسکن، مهاجرت داخلی و رشد اشتغال محلی در هلند

Why has job growth over the past decades been weaker in the Dutch Randstad area than in surrounding regions? In a simultaneous equations analysis, we find that employment has adjusted to the regional supply of labour. Net internal migration was predominantly determined by regional housing supply, and not by employment growth. Growth of the regional housing stock appeared insensitive to changes in the number of people and jobs. This lack of responsiveness to demand conditions is consistent with the presence of strong restrictions on residential development near the main Dutch cities, suggesting that the current regional distribution of economic activity in the Netherlands reflects land use planning decisions.

Government interventions in land and housing markets may have a strong impact on the quantity and location of new residential construction, while reducing the responsiveness of supply to market signals. A literature has built up that provides ample evidence of the negative impact of the stringency of land use regulations on the price elasticity of housing supply. For instance, Quigley and Raphael (2005) show that supply elasticities at the city level correlate negatively to an index of regulatory restrictiveness in California, while Green et al. (2005) and Saiz (2008) report the same relationship for a national sample of US cities. An extensive inquiry into British housing supply indicates that it is almost fully inelastic, at least partly as a consequence of the planning system (Barker, 2003 and Barker, 2004). Vermeulen and Rouwendal (2007) find that for similar reasons, housing supply in the Netherlands is almost fully inelastic as well. This literature enables us to understand the impact of land use regulation on the functioning of housing markets, but its wider effects on regional economies have received significantly less attention. Restrictions on the supply of housing that limit the number of households in a region affect labour supply and employment. For instance, Glaeser et al. (2006) and Saks (2008) show that in US cities in which such restrictions are strong, shocks in labour demand push up wages and house prices, while the local employment response is small.1 Moreover, it has been well established that the spatial distribution of jobs relates to productivity through the presence of agglomeration economies (cf. Rosenthal and Strange, 2004), so that regional productivity growth may be inhibited by restrictions on residential development too (cf. Glaeser and Gottlieb, 2008). An argument along these lines has recently been put forward by the OECD in its Territorial Review of Randstad Holland (OECD, 2007). As one of the most densely populated in the OECD, this area contains the four largest cities in the Netherlands on about 20% of all the land in this country, and its contribution to the national income presently exceeds 50%. Nevertheless, the Territorial Review points to lagging labour productivity growth in the past few years, relative to other metropolitan areas. Amongst the potential culprits, it discusses the lack of high quality dwellings, as a consequence of rigidities in Dutch housing markets. In view of the strength of government intervention in land markets, it makes sense to relate these rigidities at least partially to land use planning. Rather than consisting of one large metropolitan area, the Randstad area may be described as a ‘cluster of towns and open space’, and land use planning has consistently strived to maintain this character (Koomen et al., 2008). Key ingredients of the planning strategy are preservation of the Green Heart area between the main cities of the country (Amsterdam, Rotterdam, The Hague and Utrecht) and of so-called Buffer zones, which prevent these cities from melting together. Other important policy concepts are ‘compact cities’ and ‘clustered deconcentration’. The first of these policies aims to foster high urban densities, so that open space outside of cities is preserved, and the second aims to concentrate suburbanisation at specific locations for similar reasons. Fig. 1, which is based on Koomen et al. (2008), aptly illustrates the implementation of these policy concepts.2 It shows the contours of the present Green Heart, as well as the contours of its original conception half a century ago. When interpreting the minor shift in these contours, it should be kept in mind that Dutch population has grown by almost 50% and that the number of households has more than doubled since. Considering changes in land use between 1995 and 2003, Koomen et al. (2008) show that indeed, the propensity of transformation from open space to urban use was significantly lower in the Green Heart and the Buffer zones than in other parts of the Randstad. Moreover, Fig. 1 indicates that such changes occurred predominantly in relatively large contiguous areas that are attached to urban areas, thus adhering to the principles of compact development and clustered deconcentration. The figure also points to a significant presence of agricultural land and nature, about 75% of all land in the Randstad may be characterised as open space. Overall, these observations suggest that land use planning has succeeded in restricting urban development in this area.More evidence in support for this claim is found in the gap between residential land values at the fringe of cities and the price of agricultural land, which may be interpreted as a shadow price of restrictions on land use, or as a ‘regulatory tax’ (Glaeser et al., 2005). Using information on agricultural land prices, conversion costs and the value of residential land in a new development project at the Amsterdam fringe, Vermeulen and Rouwendal (2008) estimate that the gap amounts to roughly 30% of the average house value in Amsterdam. This exceeds regulatory tax rates for cities like Boston, New York and Washington, DC as reported in Glaeser et al. (2005). Finally, Vermeulen and Rouwendal (2007) interpret their finding that housing supply at the national level is almost fully inelastic with respect to prices in terms of restrictive planning, while ruling out several alternative explanations for the surge in real house prices, such as rising construction costs. Using a large dataset on individual housing transactions, these authors also estimate quality-controlled local house price differentials at the level of municipalities and by using information on the year of construction, they develop an index of the average attractiveness of location of different vintages of new housing supply since 1970. In spite of large spatial differences in the attractiveness of new building sites, this ‘quality of location’ index shows remarkably little variation over time. In particular, it is not the case that the more attractive locations have been developed earlier, which would have been expected in freely operating land markets. Consistent with the evidence in Fig. 1, this suggests that land use planning has not only restricted aggregate housing supply but that it has also strongly marked the spatial distribution of residential development. Motivated by the potentially significant implications that land use planning may have on productivity, our present paper investigates the extent to which housing supply has shaped the regional distribution of people and jobs in the Netherlands. We estimate three simultaneous equations for growth of the housing stock, net internal migration and employment growth on annual regional panel data that span three decades. Our econometric approach essentially follows Carlino and Mills (1987), although we extend their framework in a number of ways. First of all, we introduce an equation for growth of the housing stock as in Greenwood (1980) and Greenwood and Stock (1990). Second, as the regions in our data are not closed in terms of commuting, spatial interaction is accounted for following Boarnet (1994). Because internal migration is the main channel through which the population adjusts to regional labour and housing market conditions, we model the net internal migration rate rather than population growth (cf. Greenwood and Hunt, 1984). Moreover, the use of regional time series allows us to distinguish short-run and equilibrium adjustment effects in the interaction of our endogenous variables, while controlling fully for all national trends and time-invariant regional determinants.3 Our research question relates to a classical debate in regional science that has come to be known as the issue whether ‘people follow jobs’ or ‘jobs follow people’. A variety of studies have estimated simultaneous models for the intrametropolitan distribution of people and jobs.4 Housing supply is ignored in the larger part of this literature, which may be justified only if new construction fully accommodates demand. However, at the urban level, an upward sloping housing supply curve is implied already by the limited availability of land at a certain proximity to the city centre.5 An increase in the demand for spacious dwellings, due to rising incomes or falling transport costs for instance, will therefore push city boundaries outwards, even if all jobs remain located in a Central Business District (cf. Anas et al., 1998). So in this case, the supply of spacious dwellings drives population growth in suburbs. Simultaneous analyses of the intrametropolitan location of houses, people and jobs in the US have confirmed the empirical relevance of such mechanisms (Greenwood, 1980 and Greenwood and Stock, 1990). Our paper takes this debate to a setting where substantial restrictions on residential development near city boundaries exist. The remainder of this paper is structured as follows. Main trends in the distribution of houses, people and jobs at the more aggregate level of country parts are documented and interpreted in the next section. Section 3 introduces our data more formally and presents all variables used in the simultaneous equations model, which is estimated in Section 4. The final section concludes and offers some discussion.

Our empirical analysis identifies housing supply as a driving force behind regional development in the Netherlands. Although a strong correlation exists between regional growth of the number of houses and residents, housing supply turns out to be hardly responsive to either population or employment growth once the endogeneity of these variables is taken into account. In contrast, net internal migration appears to be highly sensitive to changes in the regional housing stock, while a growing number of jobs does not have statistically significant impact. We find that the long-run relationship between the number of people and jobs in a region is mainly restored through changes in employment growth. So regional housing supply induces population growth and in the long run, this increase in labour supply is matched by demand. The prominence of housing supply in our findings may appear surprising as, with the notable exception of Greenwood (1980) and Greenwood and Stock (1990), it is ignored in most of the empirical literature on the interdependency of local population and employment growth. However, as recently observed by Glaeser et al. (2006) and Saks (2008), the response of local labour supply to shifts in demand depends crucially on the price elasticity of housing supply. These authors show that in US cities where new construction is restricted by severe land use controls, shifts in labour demand push up house prices and wages, but employment is largely unaffected. It follows that in such cities, employment is basically determined by the size of the housing stock. Since housing supply conditions are highly restrictive in the Netherlands as well, our findings are perfectly in line with this work. Our results are also consistent with strands of the literature on internal migration and regional labour markets. Notably, the importance of housing market conditions for internal migration has been reported in various earlier studies (see Gabriel et al., 1992, for the US, Jackman and Savouri, 1992 and Cameron et al., 2006, for the UK, and Antolin and Bover, 1997, for Spain). The absence of any significant effect of employment growth on internal migration in our analysis is in line with the mixed performance of regional wage and unemployment variables in the literature (cf. Greenwood, 1993). Furthermore, labour is known to be rather immobile between regions, in particular in most European countries (cf. Eichengreen, 1993, Decressin and Fatas, 1995 and OECD, 2005). To the extent that regional labour supply adjusts to demand through internal migration, such findings suggest that the wage elasticity of regional labour supply is limited. On the other hand, the regional demand for labour should be elastic with respect to wages, in particular in a small and open economy such as the Netherlands. Although short-run elasticities are generally found to be below unity (Bartik, 1991), it seems plausible that the long-run employment response to a shift in wages is substantially larger (cf. Muth, 1990). If regional labour demand is much more sensitive to wages than supply, one should expect to find that employment adjusts to the regional distribution of people rather than the other way around. Unfortunately, the demand and supply elasticities in labour and housing markets that enable us to interpret the results in terms of underlying economic behaviour could not be estimated, because regional house prices and wages were not available.36 As a consequence of our choice to analyse annual time series spanning three decades at the regional level, the range of other explanatory variables at our disposal was also limited. While this has made the identification of causal relationships challenging, with the limited strength of instruments being a particular reason for caution in the interpretation of results, this disadvantage of our empirical strategy has been traded off against the possibility to analyse both short and long-run effects in the interdependency of our endogenous variables. Furthermore, the regional panel structure of the data has allowed us to control fully for national trends as well as for all time-invariant regional determinants of growth rates of housing, population and employment. In interpreting our results, a few other caveats should also be born in mind. In the first place, labour demand and supply are heterogeneous. Although aggregate employment has been found to adjust to the regional supply of labour, the inclination to follow jobs is likely to rise with educational attainment. If housing supply restricts the total number of workers in a booming region, higher educated workers may outbid the lower educated for housing. The existence of significant differences in educational attainment between regions in the Netherlands supports this view. In the second place, it should be realised that our results have been obtained in a setting which is characterised by restrictive land use regulation and generally tight housing market conditions. It makes sense to expect that new construction attracts workers and jobs in a region where housing supply is highly restricted. However, this finding should not be taken as a recipe for growth enhancement in lagging peripheral regions, where the size of the housing stock by and large reflects demand conditions. So why has job growth over the past decades been weaker in the Dutch Randstad area than in surrounding regions? While our analysis points to the role of lagging housing supply, it does not provide explicit evidence of the role of land use planning. However, as discussed in the introduction to this paper, there is ample evidence that land use regulation has imposed binding and significant constraints on residential development, particularly in this region. If planning restrictions would have been less severe, more houses would have been built here and our analysis suggests that in turn, this would have led to more people and jobs.37 Through foregone economies of agglomeration, these restrictions have probably left their marks on productivity too. In other words, there may be substance to the OECD claim that housing market institutions partly explain lagging labour productivity in the Randstad area.